Re: d-i on ancient hardware

[stappers@stappers.nl (Geert Stappers)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Mon, Jul 10, 2006 at 02:27:52AM +0200, Frans Pop wrote:
> On Sunday 09 July 2006 14:42, Thiemo Seufer wrote:

 [ user can't recognise an 'out of memory error'
    as an 'out of memory error' ]

> > Well, the kernel (or the bootloader?) could check that and panic.
> 
> Totally out of my league :-) I'd guess the kernel would be most logical as 
> that would cover all architectures and installation/boot methods.
> 
> If you want to follow up on that, please do.


Say we provide a 800kb kernel and a 3Mb initrd
and the user wants to load it in a computer with 2Mb RAM

Say we document that the minimal requirement are 4Mb RAM.
and the user still wants to load it in his 2Mb computer.

Freedom allows the user to try that stupid thing.

So the user comes back and says the installer dies very early
in a strange way. And we tell to the user there is NOT enough
RAM is his computer. Also that we can't detect it in the installer.

Detecting should indeed be done outside the installer:
 Kernel or bootloader.

Kernel detection:
- -----------------

I will be the last one who will be asking the kernel maintainers
to add a check for detecting not having the minimal requirement.

If I was a kernel maintainer I would have a good laugh when
seeing such a request. In a good mood I would ask if they
want also LED indicating that the battery is empty.


Bootloader:
- -----------

Imagine a bootloader that has in his 'loadloop' a check
like 
  did I reach a memory limit?
    then I have a reason to quit

I don't think the boot-loader builder will be proud that check.
It could become 
  did I reach a memory limit?
    then I have to inform the user.

Informing the user is telling the user that 3.8Mb doesn't fit
in 2Mb. The user will smile and reply
 I known, but I would try it anyway.

The boot-loader builder will leave out the memory limit check
and will be on loading what can be loaded.


So I think nothing will change in the software design
and do hope that we can learn users what minimal requirements are.


Cheers
Geert Stappers
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFEs5jKOSINbgwa/7sRAkJzAKC+dsU5G5CDnxAxNLoBrxzwrgOERgCfbjir
yXR/Dlko5AAaYiyBAA+tpmA=
=gTBo
-----END PGP SIGNATURE-----